Draft: XA: SUMMARY: AddressSanitizer: heap-use-after-free /usr/include/c++/7/bits/list.tcc:69

Problem found during RQG testing.

Thread1:

   CREATE TABLE IF NOT EXISTS t1 ( col1 INT, col2 INT, col_int INTEGER, col_int_g INTEGER GENERATED ALWAYS AS (col_int) VIRTUAL ) ENGINE = InnoDB ROW_FORMAT = Dynamic ;

  ...

   XA BEGIN 'xid1';

followed by looping with

   UPDATE t1 SET col1_copy = col1  ; 

   ALTER TABLE t1 CHANGE COLUMN IF EXISTS col1_copy col1 INT, LOCK = DEFAULT ;

Thread2:

   Looping with

    ALTER TABLE t1 ADD COLUMN col1_copy INT , LOCK = DEFAULT  ;

   ALTER TABLE t1 DROP COLUMN IF EXISTS col1, LOCK = DEFAULT  ;

The statements executed within the loops fail quite often which is IMHO to be expected.

But after rather short time comes a

==15203==ERROR: AddressSanitizer: heap-use-after-free on address 0x60300003aae0 at pc 0x564ff4abd6df bp 0x7f70c7aceed0 sp 0x7f70c7aceec0

READ of size 8 at 0x60300003aae0 thread T29

    #0 0x564ff4abd6de in std::__cxx11::_List_base<dict_v_idx_t, ut_allocator<dict_v_idx_t, true> >::_M_clear() /usr/include/c++/7/bits/list.tcc:69

#1 0x564ff4abbcb3 in std::__cxx11::_List_base<dict_v_idx_t, ut_allocator<dict_v_idx_t, true> >::~_List_base() /usr/include/c++/7/bits/stl_list.h:442

    #2 0x564ff4aaafcd in std::__cxx11::list<dict_v_idx_t, ut_allocator<dict_v_idx_t, true> >::~list() /usr/include/c++/7/bits/stl_list.h:733

    #3 0x564ff4aae3f1 in ha_innobase_inplace_ctx::~ha_innobase_inplace_ctx() (/home/mleich/Server/10.4/bld_asan/sql/mysqld+0x20203f1)

    #4 0x564ff4aae499 in ha_innobase_inplace_ctx::~ha_innobase_inplace_ctx() (/home/mleich/Server/10.4/bld_asan/sql/mysqld+0x2020499)

    #5 0x564ff3f0a1fa in Alter_inplace_info::~Alter_inplace_info() /home/mleich/Server/10.4/sql/handler.h:2461

    #6 0x564ff3effa72 in mysql_alter_table(THD*, st_mysql_const_lex_string const*, st_mysql_const_lex_string const*, HA_CREATE_INFO*, TABLE_LIST*, Alter_info*, unsigned int, st_order*, bool) /home/mleich/Server/10.4/sql/sql_table.cc:9868

    #7 0x564ff406f315 in Sql_cmd_alter_table::execute(THD*) /home/mleich/Server/10.4/sql/sql_alter.cc:490

    #8 0x564ff3cb8594 in mysql_execute_command(THD*) /home/mleich/Server/10.4/sql/sql_parse.cc:6344

    #9 0x564ff3cc4194 in mysql_parse(THD*, char*, unsigned int, Parser_state*, bool, bool) /home/mleich/Server/10.4/sql/sql_parse.cc:8154

    #10 0x564ff3c98746 in dispatch_command(enum_server_command, THD*, char*, unsigned int, bool, bool) /home/mleich/Server/10.4/sql/sql_parse.cc:1831

    #11 0x564ff3c94f61 in do_command(THD*) /home/mleich/Server/10.4/sql/sql_parse.cc:1364

    #12 0x564ff4057596 in do_handle_one_connection(CONNECT*) /home/mleich/Server/10.4/sql/sql_connect.cc:1398

    #13 0x564ff4056e51 in handle_one_connection /home/mleich/Server/10.4/sql/sql_connect.cc:1301

    #14 0x564ff55c1a94 in pfs_spawn_thread /home/mleich/Server/10.4/storage/perfschema/pfs.cc:1862

    #15 0x7f70de8f77fb in start_thread (/lib/x86_64-linux-gnu/libpthread.so.0+0x77fb)

    #16 0x7f70ddb2db5e in clone (/lib/x86_64-linux-gnu/libc.so.6+0x114b5e)

....

Observations:

1. In case I remove the XA BEGIN 'xid1';  from the actions of Thread1

    than nothing bad happens.

2. In case I rewrite the DDLs to

     ALTER .... ALGORITHM = COPY

     than nothing bad happens.